Optimized Outlier Detection and Feature Ranking for Improved ECG-CTG Signal Analysis

Main Article Content

Aditya. Y
Dr. S. Suganthi Devi
Dr. B.D.C.N Prasad

Abstract

Cardiovascular diseases (CVDs) remain the leading cause of death worldwide, making early detection crucial for reducing complications in high-risk individuals. The most common diagnostic tools for CVDs include heart sound (HS) and electrocardiogram (ECG) auscultation. These measures complement each other by evaluating the heart's electrical activity (ECG) and providing insights into its mechanical function (HS). For fetal monitoring, cardiotocography (CTG) is the most widely used non-invasive technique, as it continuously tracks the fetal heart rate (FHR) in sync with the mother’s uterine contractions (UC).Machine learning (ML) algorithms have become indispensable in medical science for detecting hidden patterns within large datasets, eliminating the need for extensive human involvement. However, ongoing research aims to further optimize these models to improve classification accuracy. Inspired by recent developments in ML models, this study proposes a heterogeneous unsupervised learning framework for real-time signal classification. The primary objective is to reduce the error rate and enhance the accuracy of a cluster-based ECG+CTG classification system. To achieve this, a novel cluster membership-based feature ranking and classification method is introduced for real-time ECG data. An enhanced Inter-Quartile Range (IQR) technique is applied to detect and remove extreme outliers, ensuring cleaner input data. Next, an ensemble feature ranking method is employed to identify key features, which are segmented for further processing. Finally, an advanced ensemble learning model is implemented to optimize the prediction accuracy across the segmented clusters. The experimental results demonstrate that the proposed approach outperforms traditional models, achieving superior performance metrics in disease prediction and classification. This framework not only improves the precision of real-time diagnostics but also sets the stage for future advancements in predictive healthcare systems.

Article Details

Section
Articles